WO2007006185A1

WO2007006185A1 - A method for protecting the channel of the voip digital signal processing

Info

Publication number: WO2007006185A1
Application number: PCT/CN2005/001814
Authority: WO
Inventors: Bin Lin
Original assignee: Zte Corporation
Priority date: 2005-07-12
Filing date: 2005-11-01
Publication date: 2007-01-18
Also published as: CN100479425C; CN1897561A

Abstract

A method for protecting the channel of the VOIP digital signal processing includes: setting a plurality of message buffer storage queues to the digital signal processing channel, and filtering the ineffective message in the said message according to the succession of the sequence numbers and/or time stamps of the real time transport protocol message arrived, and inputting the filtered message to the digital signal processing channel. The present invention overcomes the defect during the packet receiving process of the VOIP DSP chip, and solves the problem that the sound quality is abnormal and fall when the ineffective data is received in the voice channel, thereby improves the working reliability of the DSP channel, and gives VOIP equipment the high robustness and anti-interference ability.

Description

Method for protecting VOIP digital signal processing channel

The invention relates to a chip protection method for DSP (Digital Signal Processing) in the field of communication, in particular to improving the reliability and VOIP of a DSP channel by filtering invalid RTP (Real Time Transmission Protocol, for transmission media) messages. Voice over IP, IP-based voice service) The robustness of the device. Background technique

The traditional communication service adopts circuit-switching technology. The two parties establish a service by establishing a fixed-bandwidth circuit. In this way, the bandwidth utilization is low, the communication cost is high, and the development and promotion of value-added services are difficult.

At present, many newly developed multimedia communication services use packet switching technology, and the bearer network is an IP packet switching network, such as the Internet. The advantages of packet switching are high network utilization, low cost of communication, and the development and promotion of new services is relatively fast due to the use of an open architecture.

When using IP or other packet-switched networks as the bearer network, the DSP is usually used to perform PCM (Pulse Code Modulation) voice-to-packet voice bidirectional conversion. A user's voice is converted to a data packet on the packet network after being converted by the DSP. After the packet arrives at the peer, it is converted to voice by the other party's DSP, and vice versa.

When a user uses a multimedia communication service provided by an IP network, since the IP network is not as determined and reliable as the circuit-switched network, the following problems are encountered:

When an invalid packet from the network is received, the packet is also sent to the DSP because the port meets the requirements, which affects the normal operation of the DSP, resulting in a drop in voice quality or DSP anomalies.

The source of the packet may be from a network error, such as an IP or UDP port error.

It may also come from malicious cyber attacks. These packets may also be sent to the DSP due to the port's compliance requirements, causing the same problem.

In the private network environment, the device needs to periodically send media packets because it needs to maintain the internal mapping of NAT. This packet is a kind of interference to the DSP channel of the receiver and causes the same problem.

The reason is: DSP can only support one voice stream at the same time, all messages except this voice stream will have adverse effects on the DSP, and introduce delay and jitter, which will affect the voice quality. The amount, the heavy, causes software anomalies inside the DSP chip.

Chinese patent application (03125107) discloses a method for sorting by using the serial number of RTP data packets to eliminate the jitter delay thereof; the patent uses a buffer storage queue to receive RTP data packets from an IP network, and the data is entered into a queue according to the data packet number. Determining the storage address in the buffer queue and storing the data packet in its storage address, and starting the corresponding counter; when the data is dequeued, the counter corresponding to the storage address of each data packet in the buffer queue is incremented by one, Then check whether all the packets currently stored in the buffer queue meet the output condition: timer®+se _q i)=min— seq+n, if yes, check the storage address corresponding to the smallest packet sequence number. Whether there is a data packet in it, or some of the data packets are output to the next interface, and the counter corresponding to the storage address is cleared. This method uses the serial number to debounce, and cannot invalidate the miscellaneous packets. If an invalid message occurs, it is also input into the DSP.

Chinese Patent Application (03125108) discloses a method of sorting time stamps of RTP packets to eliminate their jitter delay. The method uses the buffer storage queue to receive the RTP data packet. When the data is enqueued, the first normally arrived data packet is first stored in the start address of the buffer queue, and the time stamp is used as the reference time stamp; A data packet, based on its timestamp, determines its storage address in the buffer queue and stores it in the corresponding address. When the data is dequeued, after the predetermined time is cached, the reference timestamp is used as the transmission timestamp, and the corresponding first normally arrived data packet is output to the next interface; then the sum of the transmission timestamp and the timestamp interval is used as the sum of The new transmission time stamp, and then check whether there is a data packet in the storage address corresponding to the new transmission time stamp, and then output the data packet to the next interface. This method uses the timestamp to debounce and cannot filter invalid packets. If an invalid message appears, it is also entered into the DSP. Summary of the invention

The technical problem to be solved by the present invention is to provide a method for protecting a VOIP DSP channel, which overcomes the defects in the VOIP DSP chip receiving process, and solves the problem of sound quality degradation and abnormality when the voice channel receives invalid data, thereby improving the DSP channel. The reliability of the work makes the VOIP equipment highly robust and anti-interference.

In order to achieve the above object, the present invention provides a method for protecting a VOIP digital signal processing channel, wherein a plurality of buffered message queues to the digital signal processing channel are set, and according to an incoming real-time transmission protocol message sequence The continuity of the number and/or the timestamp, filtering the invalid message in the message, and inputting the filtered message to the digital signal processing channel. The method includes the following steps:

Step 1: Set a plurality of buffered message queues to the digital signal processing channel, the queue including at least a normal queue and a protection queue;

Step 2: extract the sequence number of the newly arrived real-time transport protocol packet, and determine whether the packet is a suspicious packet according to the continuity of the sequence number and/or the timestamp of the preceding and succeeding messages; if yes, the suspicious report is The text is stored in the protection queue; if not, the message is placed in a normal queue;

Step 3: determining, according to the sequence number of the subsequent new packet, whether the suspicious packet in the protection queue is an invalid packet; if yes, discarding the suspicious packet, and clearing the protection queue; if not, using the The protection queue replaces the normal queue.

The method, further comprising the step of transmitting a message in the normal queue to a digital signal processing channel.

The method further includes setting a variable V to record a sequence number of a last message in the current normal queue.

The method further includes: determining whether the incoming message is the first message after the channel is opened, and if yes, assigning the initial value of the variable V to the sequence of the first message number.

The method further includes the step of continuously updating the variable V when the new message is stored in the normal queue.

The method, wherein the step 2 further comprises:

Step 21, set threshold values A, B;

Step 22: Determine whether the absolute value of the sequence number M of the newly arrived message and the current value of the variable V is less than the threshold A. If not, the message is a suspicious message. If yes, determining whether the timestamp of the new packet with the sequence number M and the timestamp of the old packet whose sequence number is the current value of V is less than the threshold B; if yes, The new packet is not a suspicious packet, and the new packet with the sequence number M is included in the normal queue, and the value of the variable V is updated to be M. If not, the packet is a suspicious packet.

The method, wherein the step 3 further comprises:

Step 31, setting a threshold C;

Step 32: Record the serial number L of the subsequent new message;

In the third step, it is determined whether the absolute value of the value of the variable V is less than the value of the variable V. If yes, the message stored in the protection queue is an invalid message, and the sequence number is L's new The message is a valid message and is stored in the normal queue, and the value of the variable V is updated as L; if not, the absolute value of the L is subtracted from the sequence number of the last message in the protection queue. Whether it is less than the threshold C; if yes, the newly received subsequent message with the sequence number L is also a suspicious message, and is stored in the protection queue. If not, the The packet in the protection queue is an invalid packet, and the protection queue is emptied, and the packet with the sequence number L is stored as a suspicious packet in the protection queue.

The method, wherein the step three or three further comprises:

Step 3: 31, setting a threshold value of the number of protection queue packets D;

In step 3:32, it is determined whether the number of the packets stored in the protection queue is greater than or equal to the threshold D. If yes, the original normal queue is cleared, and the original queue is replaced by the protection queue to form a new one. a normal queue, and further updating the value of the variable V as the serial number of the last message in the new normal queue; if not, the V value is unchanged, continue to extract the serial number of the new message and transfer The method of the step, wherein the step 323 further includes: when the number of packets in the protection queue is greater than or equal to the threshold D, using the original normal queue as a new protection queue. A step of.

The invention solves the defects in the VOIP DSP packet processing in the prior art, and avoids the degradation of the sound quality and the abnormality when the invalid data is received. When the received packet is normal, it does not affect the original packet processing process, and does not introduce packet loss, jitter, and delay. When there are invalid messages in the received message, these invalid messages can be filtered out to avoid abnormalities, improve sound quality, and not introduce new packet loss, jitter, and delay. When media/call switching is performed, a delay of several tens of milliseconds is introduced at the time of switching, which can be ignored, but an abnormality can be avoided and the sound quality is improved.

This method can be applied to all communication systems that use RTP as a media transport protocol. There is no need to modify existing call control protocols. BRIEF DESCRIPTION OF THE DRAWINGS - Figure 1 is a process flow diagram of the method of the present invention;

2 is a sequence diagram of an RTP message entering a DSP before applying the method of the present invention;

3 is a sequence diagram of an RTP message entering a DSP after applying the method of the present invention. detailed description:

The normal RTP packets are continuous. Occasionally, due to out-of-order and packet loss, they fluctuate within a small range. When media switching occurs (refers to the switching of normal voice RTP streams), the sequence number jumps, but subsequent The message should be continuous again. According to these features, the suspicious packet can be temporarily placed in the protection queue. If the subsequent packet proves that the suspicious packet is invalid, the temporary queue is cleared; if the subsequent packet proves that the suspicious packet is indeed valid Then, replace the normal queue with the protection queue and pass the data to the DSP. The original normal queue acts as a new protection queue. The method of the present invention can be implemented by judging the continuity of the message sequence number or the continuity of the time stamp, or in combination, by simultaneously determining the continuity of the sequence number and the time stamp.

Figure 1 shows a process flow diagram of the method of the invention, the method of the invention being as follows:

1) Each DSP channel is assigned a number of buffered message queues, including the normal queue and the protection queue. The variable is reassigned to record the sequence number of the last message entering the current normal queue, assuming that a variable named V is assigned.

2) When the DSP receives a new RTP message (step 101), it is determined whether the message is the first message after the channel is opened (step 102), and if so, the message is placed in the normal queue. That is, when the DSP receives the first message, put the message into the normal queue, and use the variable V to record the serial number of the message. Assuming that the sequence number of the message is N, then V = N (step 103);

3) If no, extract the sequence number of the message (assumed to be equal to M) (step 104), determine whether M = v + l (step 105), if M = v + 1, then consider the RTP message to be a A completely normal and continuous message, put this message into the normal queue, and record the new serial number with the variable V, ie v = M (step 106).

4) It is judged that if the absolute value of M and V subtraction is not large, that is, the absolute value of M minus V is less than a threshold A (this threshold is configurable, for example, set to 10), and the serial number is M, V. The absolute value of the time stamp subtraction of the message is less than a threshold B (this threshold is also configurable, for example, set to 100 milliseconds) (step 107), then the message is considered normal, possibly due to The lower layer is unreliable transmission, and out of order or packet loss occurs. Put this message in the normal queue. Write down the new serial number with the variable V, ie v = M (step 108).

5) If the absolute value of M and V subtraction is relatively large (ie, the absolute value of M minus V is greater than or equal to the threshold A), or the absolute value of the time stamp subtracted by the message with the sequence number M and V is greater than or equal to Threshold B, consider this message suspicious, put this message in the protection queue. The value of the variable V does not change. Wait The subsequent message is determined to be processed in the next step (step 109).

6) When receiving the subsequent RTP message, extract the serial number of the message (assumed to be equal to L) (step 110), and determine whether the difference between L and variable V is not large (ie, the absolute value of L minus V) If it is less than the threshold A) (step 111), if yes, it is considered that the suspected packet in the protection queue is an invalid packet, clear the protection queue, put the packet with the serial number L into the normal queue, and record the new variable with the variable V. The serial number, ie v = L (step 112).

7) If the difference between L and variable V is large, that is, the absolute value of L minus V is greater than the threshold A, it is further determined whether the message with the sequence number L is continuous with the message in the protection queue, that is, L minus Whether the absolute value of the sequence number of the last message in the protection queue is less than the threshold C (step 113). If no, that is, the new message is less continuous than the message in the protection queue, it is considered that not only the previous one is The suspected packet is an invalid packet, and the newly received packet is also suspicious. The protection queue is first cleared, and the newly received packet with the serial number L is placed in the protection queue (step 114). Note that the threshold C and the threshold A can be unequally configured separately. For example, the threshold C can be configured to 5.

8) If L and variable V are significantly different (ie, the absolute value of L minus V is greater than the threshold A), but compared with the message in the protection queue is relatively continuous (L minus the last one of the protection queue) The absolute value of the sequence number of the message is less than the threshold C), and the message with the sequence number L is placed in the protection queue (step 115). Check the number of packets in the protection queue to determine whether it is less than the threshold D (step 116). If the number is less than the threshold D (this threshold is also configurable, for example, configured as 3), the variable V The value does not change, wait for the next message (step 118), and returns to step 110 to wait for subsequent messages to determine the next step. Until the number of packets in the protection queue is greater than or equal to the threshold D, the process proceeds to step 117.

9) At this time, it is considered that the existing suspected packets in the protection queue are valid packets, and the original normal queue is replaced by the protection queue as the new normal queue, and the original normal queue is cleared and used as a new protection queue. Use the variable V to note the serial number of the last message number received (assuming the sequence number is X), ie v = X (step 117).

Among them, the thresholds A, B, and C are all configurable in the above steps. In addition, when the number of consecutive RTP packets received by the protection queue is upgraded to a normal queue, it can be configured, that is, the threshold D is also configurable. The larger the value of the threshold value D, the smaller the possibility of false switching, but the larger the switching delay. In engineering, you can find an optimal value by flexibly configuring these thresholds, so that the overall performance of voice quality, handover delay, and device robustness is optimal. As an application example, other applications are similar. 2 is a sequence diagram of an RTP message entering the DSP before applying the method of the present invention. 3 is a sequence diagram of an RTP message entering a DSP after applying the method of the present invention.

Here, two queues are assigned as an example, one as a normal queue and one as a protection queue. In this example, threshold A is configured to 10, threshold B is configured to 100ms, threshold C is configured to 5, and threshold D is configured to 2.

In the first step, the packet with sequence number 1 is placed in the normal queue (the initial normal queue 31 shown in Figure 3), waiting for the DSP driver to input the message into the DSP. Record current serial number = 1.

In the second step, the packet with sequence number 2 is received in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 2.

The third step is to receive the packet with the sequence number 4. The sequence number of the packet is not continuous, but the difference is not large (less than the threshold A). At this time, the timestamp is checked, and the timestamp is found to be similar (less than the threshold B). Place it in the normal queue and wait for the DSP driver to enter the message into the DSP. Record current serial number = 4.

The fourth step, receiving the packet with the serial number of 1000, the serial number of the packet is completely discontinuous (greater than the threshold A), indicating that the packet is suspicious. Since it is currently impossible to determine whether the packet is invalid or media switching occurs, so this is The message is placed in the protection queue, the initial protection queue 33 shown in FIG.

In the fifth step, the packet with the sequence number 5 is received, and the sequence number of the packet is consecutive with the current sequence number of the normal queue. At this time, it can be determined that the suspected packet in the protection queue is an invalid packet, and the 1000 packet in the protection queue is cleared, and the serial number is The 5 packet is placed in the normal queue, waiting for the DSP driver to input the message into the DSP. Record the current serial number - 5.

In the sixth step, the packet with sequence number 6 is received in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 6.

In the seventh step, the packet with the sequence number 1001 is received, and the sequence number of the packet is completely discontinuous (greater than the threshold A), indicating that the packet is suspicious. Since it is currently impossible to determine whether the packet is an invalid message or a media switch occurs, this is The message is placed in the protection queue.

In the eighth step, the packet with the sequence number 1002 is received, and the sequence number of the packet and the current sequence number of the normal queue are not consecutive, but compared with the 1001 message in the protection queue is continuous (less than the threshold C), the packet is first placed. In the protection queue, check whether the number of packets in the protection queue reaches the threshold D. Since the threshold D is now set to 2, the threshold has been reached. At this time, it can be determined that the media switching has occurred. The protection queue is marked as a normal queue, which is the switched normal queue 32 shown in FIG. 3, and the original normal queue is cleared at the same time, and is marked as a protection queue for use. Wait for the DSP driver to input the 1001 and 1002 messages into the DSP. Record current serial number = 1002.

In the ninth step, the packet with the sequence number 1003 is received in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 1003.

In the tenth step, the packet with the sequence number 10 is received, and the sequence number of the packet is discontinuous (greater than the threshold A), indicating that the packet is suspicious. Since the packet cannot be determined to be an invalid message or a media switch occurs, the report is sent. The text is placed in the protection queue, the switched protection queue 34 shown in FIG.

In the eleventh step, the packet with the sequence number 1005 is received, and the sequence number of the packet is very close to the current sequence number of the normal queue (less than the threshold A). At this time, it can be determined that the suspected packet in the protection queue is an invalid packet, and the protection queue is cleared. In the 10th package, the packet with the serial number of 1005 is placed in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 1005.

In the twelfth step, the packet with the sequence number 1006 is received and placed in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 1006.

In the thirteenth step, the packet with the sequence number 1007 is received and placed in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 1007.

In the fourteenth step, the packet with the serial number of 2000 is received, and the sequence number of the packet is completely discontinuous (greater than the threshold A), indicating that the packet is suspicious. Since it is currently impossible to determine whether the packet is an invalid message or a media switch occurs, This message is placed in the protection queue.

In the fifteenth step, the packet with the serial number of 3000 is received, the sequence number of the packet is not continuous with the normal queue (greater than the threshold A), and the packet in the protection queue is not continuous (greater than the threshold C:), indicating that the packet is suspicious and cannot be Determine whether the message is invalid or media switch, delete the 2000 packet in the protection queue, and put the packet with the serial number 3000 into the protection queue.

In the sixteenth step, the packet with the sequence number 1008 is received, and the sequence number of the packet is consecutive with the current sequence number of the normal queue. At this time, it can be determined that the suspected packet in the protection queue is an invalid packet, and the protection queue is cleared. Put the 1008 packet in the normal queue and wait for the DSP driver to input the message into the DSP. Record the current serial number = 1008.

In the seventeenth step, the packet with the sequence number 1009 is received and placed in the normal queue, waiting for the DSP driver to input the message into the DSP. Record current serial number = 1009.

After this process, refer to Figure 2 to apply the method before the RTP message enters the DSP sequence diagram and attached Figure 3 is a sequence diagram of the RTP message entering the DSP after applying the method. Figure 2 uses the prior art single-queue mode. In this way, all messages enter the DSP directly. Figure 3 uses the dual queue mode to set the normal queue and the protection queue respectively. Only the packets in the normal queue enter the DSP. It can be seen from the comparison before and after the method of the present invention that only the packets in the normal queue enter the DSP, and the original RTP packets with many interferences are processed into a very regular, very "clean" voice. The RTP stream is input to the DSP to protect the DSP.

The conditions for judging valid messages in the above process can be flexibly configured, for example: Threshold A, Threshold B, Threshold C, Threshold D. The above conditions can be combined as appropriate and flexibly modified. Achieve the best value for the project. Industrial applicability

The invention solves the defects in the VOIP DSP packet processing in the prior art, and avoids the degradation of the sound quality and the abnormality when the invalid data is received. When the received packet is normal, it does not affect the original packet processing process, and does not introduce packet loss, jitter, and delay. When there are invalid messages in the received message, these invalid messages can be filtered out to avoid abnormalities, improve sound quality, and not introduce new packet loss, jitter, and delay. When media/call switching is performed, a delay of several tens of milliseconds is introduced at the time of switching, which can be ignored, but an abnormality can be avoided and the sound quality is improved. This method can be applied to all communication systems that use RT as a media transport protocol. There is no need to modify existing call control protocols.

Claims

Claim

A method for protecting a VOIP digital signal processing channel, characterized by: setting a plurality of buffered message queues to the digital signal processing channel, and according to an incoming real-time transmission protocol message sequence number and/or time The continuity of the stamp, filtering the invalid message in the message, and inputting the filtered message to the digital signal processing channel.

The method according to claim 1, further comprising the following steps: Step 1: Set a plurality of buffered storage message queues to the digital signal processing channel, the queue including at least a normal queue and a protection queue;

3. The method of claim 2, further comprising the step of transmitting a message in said normal queue to a digital signal processing channel.

4. The method of claim 2 or 3, further comprising setting a variable V to record the sequence number of the last message in the current normal queue.

The method according to claim 4, further comprising determining whether the incoming message is the first message after the channel is opened, and if so, assigning the initial value of the variable V to The serial number of the first message.

6. The method of claim 5, further comprising the step of continuously updating the variable V when the new message is stored in the normal queue.

7. The method of claim 2, 3, 5 or 6, wherein the step two further comprises:

Step 21, set threshold values A, B;

Step 22: Determine whether the absolute value of the sequence number M of the newly arrived message and the current value of the variable V is less than the threshold A. If not, the message is a suspicious message. If yes, further determining the timestamp of the new packet with the sequence number M and the sequence number being the current value of V Whether the time stamp of the old message is less than the threshold B; if yes, the new message is not a suspicious message, the new message with the sequence number M is included in the normal queue, and the variable V is updated. The value is M. If no, the message is a suspicious message.

The method according to claim 7, wherein the step 3 further comprises: step three, setting a threshold C;

Step 32: Record the serial number L of the subsequent new message;

In the third step, it is determined whether the absolute value of the value of the variable V is less than the value of the variable V. If yes, the message stored in the protection queue is an invalid message, and the sequence number is The new message of L is stored in the normal queue as a valid message, and the value of the variable V is updated to be L; if not, the L is further determined to be subtracted from the sequence number of the last message in the protection queue. If the absolute value is less than the threshold C; if yes, the newly received subsequent message whose sequence number is L is also a suspicious message, and is stored in the protection queue. The packet stored in the protection queue is invalid, the protection queue is emptied, and the packet with the sequence number L is stored as a suspicious message in the protection queue.

The method according to claim 8, wherein the step three three further comprises: step three three one, setting a threshold D of the number of protection queue packets;

In step 3:32, it is determined whether the number of the packets stored in the protection queue is greater than or equal to the threshold D. If yes, the original normal queue is cleared, and the original queue is replaced by the protection queue to form a new one. a normal queue, and further updating the value of the variable V as the serial number of the last message in the new normal queue; if not, the V value is unchanged, continue to extract the serial number of the new message and transfer Step

The method according to claim 9, wherein the step 323 further comprises: when the number of packets in the protection queue is greater than or equal to the threshold D, the original The normal queue acts as a new protection queue step.